The main function is drgee, which estimates a parameter
\(\beta\) in a model defined as
\(g\{E(Y|A,L)\}-g\{E(Y|A=0,L)\}=\beta^T \{A\cdot X(L)\}\). By supplying (nuisance) models for
\(g\{E(Y|A=0,L)\)
and \(E(A|L)\), a consistent estimate of \(\beta\) is
obtained when at least one of these models is correctly specified.
In the function drgee, three estimation methods are
implemented:
O-estimation, where a model for \(g\{E(Y|A=0,L)\) is used;
E-estimation, where a model for \(E(A|L)\) is used; and
DR-estimation where models for both \(g\{E(Y|A=0,L)\) and
\(E(A|L)\) are used.
The function gee is an implementation of standard GEE with
independent working correlation matrix.
For conditional methods with clustered data, cluster-specific intercepts are assumed in all models.
The function drgeeData is used for extraction and manipulation
of data, and is called by drgee and gee.
The function RobVcov is used to calculate standard errors of
the estimates, given a vector of the residuals from estimating
equations, the Jacobian, and a cluster-identifying variable.
The function findRoots solves a system of non linear equations.